Skip to content
AI Security

Interview question · Adversarial ML & Model Privacy

A deployed model may be leaking its training data. How do you respond?

The leak is in the weights, so the playbook differs from a database breach. The scored answer confirms, contains at the output, scopes from data lineage, and decides between filtering, retraining and withdrawal.

hardNewPreparation tags: OpenAI, Google DeepMind, AnthropicUpdated Nov 2026

The leak is in the weights, so the playbook differs from a database breach. The scored answer confirms, contains at the output, scopes from data lineage, and decides between filtering, retraining and withdrawal.

Make your first attempt

What would you say before reading the answer?

  1. 01 State the assumptions your answer depends on.
  2. 02 Explain the mechanism and choose a control.
  3. 03 Name a condition that would change your decision.

The worked answer includes an approach, scoring criteria and follow-up questions. Compare the reasoning with your own.

10 answers per topic instead of 5, plus saved progress and bookmarks · costs nothingor unlock all 529 remaining answers · $25

Every worked answer in all 17 topics, every lesson in every course, the full concept curriculum, the complete practice-test bank, and the downloadable security field guides.

6 months · One payment · No auto-renewal

What if · the follow-ups in this answer

The answer works through 6 follow-ups like an interviewer would

A user reports that your public assistant output a real person's phone number and street address in response to an unrelated question.

  1. What do you confirm first?
  2. How do you contain while fixing?
  3. How do you scope the incident?
  4. What if the leaked data includes secrets, such as API keys?
  5. What is the lasting fix?
  6. How would you run the incident end to end?

Each rung has a nudge, a model answer, the principle that carries over and the weak answer to avoid. Premium opens every answer and ladder.

UP NEXT ON YOUR JOURNEY
Next in this trackWhat can watermarking of model outputs or weights prove, and what can it not?Next in this trackWhy is membership inference hard even to define for LLMs trained on internet-scale data?Next in this trackWhy can adversarially robust models leak more membership information about their training data?
DISCUSSION · 0

No comments yet. Share a correction or a different way to answer it.